ISPRS Open Journal of Photogrammetry and Remote Sensing最新文献

{"title":"Through-water and underwater photogrammetry for fossil exposure survey in shallow waters","authors":"Lucas Moreira Furlan,&nbsp;Eduardo Guilherme Piazentim,&nbsp;Lucas Henrique Garcia-Gonzalez,&nbsp;Lucas Verissimo Warren,&nbsp;Gottfried Mandlburger","doi":"10.1016/j.ophoto.2026.100143","DOIUrl":"10.1016/j.ophoto.2026.100143","url":null,"abstract":"<div><div>Subaqueous fossiliferous outcrops constitute valuable yet poorly documented paleontological archives, particularly in fluvial environments where accessibility, hydrodynamic reworking, and rapid surface degradation hinder systematic investigation. This study introduces an application of through-water (multimedia) and underwater photogrammetry for the <em>in situ</em> documentation of fossil exposures in optically shallow water (OSW) settings. The approach is applied to a permanently submerged fossiliferous pavement of the Permian <em>Irati</em> Formation exposed on the riverbed of the Passa Cinco River, southeastern Brazil, which preserves exceptionally well-preserved remains of the aquatic reptile <em>Mesosaurus</em>. Drone-based through-water photogrammetry was used to generate high-resolution orthophotos and refraction-corrected digital elevation models at the outcrop scale, enabling detailed spatial mapping of fossil occurrences and geological features. In parallel, close-range underwater photogrammetry acquired through a wading-based workflow provided high-resolution three-dimensional reconstructions of selected specimens, preserving fine-scale morphological and taphonomic attributes. The results demonstrate that the combined workflow enables qualitative and semi-quantitative, non-invasive paleontological investigations in shallow aquatic environments and highlights the potential of integrated photogrammetry for the long-term conservation and reproducibility of submerged fossil records. Furthermore, the proposed workflow enabled effective taphonomic analysis and digital preservation of fragile and dynamically exposed fossiliferous surfaces. All orthophoto datasets and dense point cloud data generated in this study are openly available in a public repository, ensuring data accessibility, reproducibility, and long-term reuse.</div></div>","PeriodicalId":100730,"journal":{"name":"ISPRS Open Journal of Photogrammetry and Remote Sensing","volume":"21 ","pages":"Article 100143"},"PeriodicalIF":7.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148738851","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Evaluation of lab-based, hybrid, and in-situ radiometric calibration for TLS","authors":"Helena Laasch,&nbsp;Tomislav Medic,&nbsp;Andreas Wieser","doi":"10.1016/j.ophoto.2026.100137","DOIUrl":"10.1016/j.ophoto.2026.100137","url":null,"abstract":"<div><div>Radiometric compensation is essential for obtaining accurate reflectance values from terrestrial laser scanning (TLS) intensities, yet it remains unclear which strategy best compensates for the impact of measurement configuration (angle of incidence (AOI), and distance) in terms of precision and accuracy. Herein, precision refers to the consistency of compensated intensities within a dataset, while accuracy describes their agreement with true reflectance values. Whether precision or accuracy should be prioritized depends on the downstream application: segmentation typically benefits from high precision, whereas material probing requires high accuracy. Compared to the distance effect, the AOI effect is harder to model because it is inherently material-dependent. Consequently, this paper places a particular focus on optimizing AOI compensation. We compare three calibration strategies: lab-based, in-situ, and hybrid. We extend our analysis by i) comparing global and material-dependent AOI compensation, i.e., using single representative vs. multiple locally adapted compensation functions, and ii) assessing the impact of the choice of laboratory calibration targets. We find that all calibration strategies yield similar precision and accuracy. The choice of strategy is thus primarily driven by operational constraints and convenience (e.g., availability of lab facilities vs. in-situ reference targets). Moreover, we show that the material-dependent AOI compensation increases precision by approximately 10%, while yielding similar accuracy as using a common (global) AOI compensation function. All three strategies lead to absolute reflectances that agree with spectrometer measurements (reference) within about 5% (median of differences). Our results confirm the potential of using TLS intensities for material probing and can serve as a guideline for choosing the TLS radiometric compensation strategy providing the best trade-off in terms of adaptability, generalizability, precision and accuracy across different datasets acquired by a common instrument.</div></div>","PeriodicalId":100730,"journal":{"name":"ISPRS Open Journal of Photogrammetry and Remote Sensing","volume":"21 ","pages":"Article 100137"},"PeriodicalIF":7.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148737386","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Disentangling geometry and texture: An investigation into large-scale RGB-DSM pre-training for 3D urban understanding","authors":"Guneet Mutreja,&nbsp;Ksenia Bittner","doi":"10.1016/j.ophoto.2026.100139","DOIUrl":"10.1016/j.ophoto.2026.100139","url":null,"abstract":"<div><div>Recent Earth observation foundation models have achieved strong transfer performance using large-scale self-supervised pre-training, but most remain primarily spectral or RGB-based and do not explicitly exploit geometric information from digital surface models (DSMs). This limits their effectiveness in dense urban scenes, where height, building boundaries, and vertical structure are essential for tasks such as building extraction and height estimation.</div><div>In this work, we present <span>HiRes-FusedMIM</span>, an empirical study of large-scale multi-modal pre-training for high-resolution urban remote sensing. We curate a dataset of 1.2 million paired RGB and DSM image tiles from native source products primarily between 0.1 to 0.5<!--> <!-->m spatial resolution, processed to training resolutions of 0.2 to 0.5<!--> <!-->m where appropriate, making it, to our knowledge, one of the largest publicly available high-resolution paired RGB–DSM datasets. Specifically, we compare shared and modality-specialized dual encoders, masked image modeling and contrastive alignment, and small- versus large-scale pre-training.</div><div>Our results show that architecture is critical for geometry-sensitive transfer. A shared encoder suffers from modality interference and drops to 18.0 AP50 on UBCv2 instance segmentation, whereas the dual encoder reaches 31.5 AP50. Similarly, the dual encoder improves LoveDA semantic segmentation from 41.58 to 52.28 mIoU and reduces height-estimation error from 9.26 m to 7.89 m RMSE. We further find that global contrastive alignment benefits some scene classification tasks but degrades dense geometric prediction. Finally, scaling unique training samples from 368k to 1.2M improves UBCv2 AP50 by 3.0 points, indicating that data diversity is more effective than repeated training on a smaller corpus. These findings provide practical guidance for designing geometry-aware foundation models for urban remote sensing.</div></div>","PeriodicalId":100730,"journal":{"name":"ISPRS Open Journal of Photogrammetry and Remote Sensing","volume":"21 ","pages":"Article 100139"},"PeriodicalIF":7.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148776822","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enhancing macrophyte sub-classification through neighborhood-independent full-waveform LiDAR","authors":"Jan Rhomberg-Kauert,&nbsp;Lucas Dammert,&nbsp;Laure-Anne Gueguen,&nbsp;Florian Pöppl,&nbsp;Michael Grömer,&nbsp;Martin Pfennigbauer,&nbsp;Gottfried Mandlburger","doi":"10.1016/j.ophoto.2026.100144","DOIUrl":"10.1016/j.ophoto.2026.100144","url":null,"abstract":"<div><div>In the clear, shallow waters of the littoral zone, bathymetric LiDAR, SONAR, and photogrammetry have proven suitable for mapping both the underwater landscape and vegetation. In the past, all three methods have been used to map underwater vegetation, but automated workflows for sub-classification of vegetation remain limited. In the field of forestry such methods, e.g. classification based on the echo ratio, have long been used to classify vegetation based on neighborhood information. However, these methods developed for terrestrial landscapes have not been applied to aquatic vegetation. Therefore, this study adapts the echo ratio for the (neighborhood) sub-classification of submerged macrophytes and evaluates the results across the three surveying techniques. Furthermore, making use of the recent advances in full-waveform LiDAR, a neighborhood-independent (waveform-based) sub-classification is introduced. The waveform-based sub-classification approximates and subtracts the water column backscatter for each recorded waveform, extracting a residual waveform that contains only vegetation echoes. The residual waveform is then analyzed for the distance between the peaks in relation to the residual length. From this, the waveform echo ratio is derived and then compared to the neighborhood echo ratio. Both sub-classification methods are additionally validated using manually annotated data. The waveform data, LiDAR data, and reference (Photogrammetric + SONAR) data produce F1-scores of 0.19/0.35/0.32 for high vegetation, 0.18/0.23/0.24 for medium vegetation, and 0.80/0.85/0.85 for low vegetation. Although these scores vary, so do those of the manual annotation, which achieved median F1-scores of 0.13/0.34/0.80 compared to each other.</div></div>","PeriodicalId":100730,"journal":{"name":"ISPRS Open Journal of Photogrammetry and Remote Sensing","volume":"21 ","pages":"Article 100144"},"PeriodicalIF":7.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148737385","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multispectral airborne laser scanning dataset for tree species classification: MS-ALS-SPECIES","authors":"Matti Hyyppä,&nbsp;Klaara Salolahti,&nbsp;Eric Hyyppä,&nbsp;Xiaowei Yu,&nbsp;Josef Taher,&nbsp;Leena Matikainen,&nbsp;Matti Lehtomäki,&nbsp;Paula Litkey,&nbsp;Teemu Hakala,&nbsp;Harri Kaartinen,&nbsp;Juha Hyyppä,&nbsp;Antero Kukko","doi":"10.1016/j.ophoto.2026.100142","DOIUrl":"10.1016/j.ophoto.2026.100142","url":null,"abstract":"<div><div>The shift from stand-level to individual-tree-level forest assessments supports improved species mapping and biodiversity monitoring, particularly in boreal ecosystems where tree species like aspen (<em>Populus tremula</em> L.) play a keystone role. Airborne laser scanning (ALS) is the standard for such inventories, but a major limitation for developing improved species classification methods is the small number of publicly available ALS datasets containing high-quality, field-validated reference data. Recently, multispectral ALS data has shown promise for tree species classification, but the progress is hindered by the lack of open multispectral ALS datasets with high-quality field reference data. This paper presents and details an open multispectral ALS dataset for tree species classification that was used before its public release for an international benchmarking study of machine learning and deep learning classification methods in a related publication by Taher et al., (2026). The dataset comprises 6326 segment-level point clouds of individual trees representing nine species in southern Finland. The point cloud data has been acquired using two multispectral laser scanning systems each operating at three laser wavelengths: a helicopter-borne system (HeliALS) with a point density exceeding 1000 <span><math><mrow><mi>points</mi><mo>/</mo><msup><mrow><mi>m</mi></mrow><mrow><mn>2</mn></mrow></msup></mrow></math></span> and an Optech Titan system with approximately 35 <span><math><mrow><mi>points</mi><mo>/</mo><msup><mrow><mi>m</mi></mrow><mrow><mn>2</mn></mrow></msup></mrow></math></span>. Furthermore, we present a crowdsourcing application that facilitates the collection of high-quality field reference data of tree species in an efficient and scalable manner. Our article showcases the versatility of the open dataset by presenting new analyses on species classification using multispectral data building upon the initial findings of Taher et al., (2026). We offer an improved benchmark of classification methods by evaluating the classification accuracy exclusively on tree segments with minimal segmentation errors, which is shown to improve the classification accuracy by 2–3 percentage points. The ranking of the classification methods remains largely unchanged and the point transformer model still achieves the highest accuracy, thus confirming the results of the prior benchmarking study. Furthermore, we study the relation between classification accuracy and tree height to demonstrate the advantage of the point transformer model for small trees and minority species. The public release of this comprehensive dataset and the related methodology facilitate the development of deep learning models for accurate tree species classification using multispectral laser scanning, while also advancing research in ecological monitoring and sustainable forest management.</div></div>","PeriodicalId":100730,"journal":{"name":"ISPRS Open Journal of Photogrammetry and Remote Sensing","volume":"21 ","pages":"Article 100142"},"PeriodicalIF":7.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148776821","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enabling aerial localization and object detection through monocular height estimation: A hybrid deep learning solution","authors":"Mohammad Momeni Esfahani,&nbsp;Mahmod Reza Sahebi,&nbsp;Mehdi Mokhtarzade","doi":"10.1016/j.ophoto.2026.100141","DOIUrl":"10.1016/j.ophoto.2026.100141","url":null,"abstract":"<div><div>Accurate Digital Surface Models (DSMs) are fundamental to critical urban applications, including flood risk assessment, communication network line-of-sight planning, and 3D city modeling. However, traditional DSM generation relies heavily on expensive airborne LiDAR campaigns or multi-view stereo imagery, which suffer from high acquisition costs, sparse global coverage, and long revisit times. Monocular Height Estimation (MHE) from single very high-resolution (VHR) remote sensing images offers a transformative, cost-effective alternative, yet its practical adoption has been constrained by difficulties in handling extreme object-scale variations and modeling long-range spatial dependencies in complex urban scenes. To overcome these barriers, this study introduces a novel hybrid deep learning framework for MHE. The presented encoder–decoder architecture features a new multi-scale feature extraction module and integrates self-attention mechanisms to improve spatial reasoning, complemented by a hybrid decoder and a self-attention auxiliary refinement head. Critically, tangible application value is demonstrated through the coupling of the proposed MHE framework with a robust template matching algorithm, enabling precise object and aerial platform detection capabilities that are directly deployable for emergency disaster response, autonomous drone navigation, and rapid updating of outdated geospatial databases in infrastructure-poor regions. Moreover, in recognition of the operational constraints of real-world deployment, a lightweight variant, Tiny MHE, is proposed, designed for efficient inference on edge. Extensive evaluations on three datasets, including two low-tilt benchmarks, show that our framework substantially outperforms current state-of-the-art methods, achieving RMSE of 1.06 m on Vaihingen and 1.582 m on Bojnord, thereby confirming its strong potential for high-impact, large-scale geospatial applications.</div></div>","PeriodicalId":100730,"journal":{"name":"ISPRS Open Journal of Photogrammetry and Remote Sensing","volume":"21 ","pages":"Article 100141"},"PeriodicalIF":7.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148738850","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Transferring instance and semantic segmentation across terrestrial laser scanning time series: towards automated forest monitoring plot measurements","authors":"E. Honkanen,&nbsp;T. Yrttimaa,&nbsp;L. Terryn,&nbsp;T. Kamula,&nbsp;A. Erkkilä,&nbsp;L. Liikonen,&nbsp;J. Hyyppä,&nbsp;M. Vastaranta","doi":"10.1016/j.ophoto.2026.100147","DOIUrl":"10.1016/j.ophoto.2026.100147","url":null,"abstract":"<div><div>Monitoring forest structural changes over time is important for sustainable management and ecological research. This requires repeated measurements on site, increasingly obtained using terrestrial laser scanning (TLS). However, conventional TLS-based methods are computationally intensive and not tailored for monitoring, as they do not exploit prior measurements. To improve the accuracy of forest monitoring and increase automation in data analyses, we developed a temporally consistent workflow, Tree Skeleton Transfer. This approach enables the transferring of both instance (tree ID labeling) and semantic segmentation (point class labeling) across TLS time series using skeletonized tree structures as stable spatial references. Using this method, instance and semantic labels are propagated to each point in a new, unclassified point cloud via a nearest neighbor assignment to the structural tree skeleton, which encodes label information from the initial segmentation of existing TLS data. The initial segmentation is performed in two stages: first, the sample plot point cloud is automatically segmented, and second, this segmentation is manually inspected and corrected if necessary. Subsequent point clouds are then segmented automatically, leveraging structural tree skeletons. Because changes in trees’ structures are often gradual, stem and leaf points are likely to be found in roughly the same locations over time. We applied this method to nine boreal forest plots in Evo, Southern Finland, using TLS data from 2014, 2023, and 2024. By leveraging semi-automated initial segmentations from 2023, tree ID and stem/crown labeling were transferred to unclassified 2014- and 2024-point clouds. The method achieved high instance segmentation accuracy, with point-level F1-scores of 0.889 when transferring labels backward to 2014, and 0.959 when transferring forward to 2024. For comparison, we adopted RayCloudTools as an alternative, automatic segmentation method applied in each data independently, achieving F1-scores of 0.791 for the 2014 data and 0.846 for the 2023 data. Semantic segmentation was robust across transfer directions, with F1-scores of 0.96–0.97 for stem points and 0.88–0.89 for crown points, exceeding the performance of an independent automated processing with F1-scores of 0.57–0.59. Compared to independent automated processing of subsequent datasets, the Tree Skeleton Transfer approach reduced computation times by 96–97% while enhancing segmentation consistency, offering clear benefits for monitoring forest structural changes over time. These findings suggest that skeleton-based segmentation transfer provides a scalable solution for operational forest inventories, with potential applications in other laser scanning contexts.</div></div>","PeriodicalId":100730,"journal":{"name":"ISPRS Open Journal of Photogrammetry and Remote Sensing","volume":"21 ","pages":"Article 100147"},"PeriodicalIF":7.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148776820","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Topography-aware registration refinement of Landsat 8/9 L1TP products with Sentinel-2 L1C and NAIP orthoimages","authors":"Jiageng Zhong,&nbsp;Shuang Li,&nbsp;Xiangxi Tian,&nbsp;Rajagopalan Rengarajan,&nbsp;Aparajithan Sampath,&nbsp;Cody Anderson,&nbsp;Jie Shan","doi":"10.1016/j.ophoto.2026.100136","DOIUrl":"10.1016/j.ophoto.2026.100136","url":null,"abstract":"<div><div>Satellite orthoimage products are widely available, yet residual geometric misregistration across independently processed products remains common, limiting their direct interoperability for multi-source and multi-temporal analysis. This paper introduces a residual Multi-Layer Perceptron (MLP)–based, topography-aware framework to further increase the registration quality of existing orthoimages based on dense matching without requiring any sensor-specific metadata or physical imaging parameters. We evaluate our approach by precisely co-registering 26 Landsat 8/9 Collection 2 Level-1 Terrain Precision (L1TP) scenes across Indiana (USA) and Berlin (Germany) with Sentinel-2 L1C and National Agriculture Imagery Program (NAIP) orthoimages. The baseline Landsat misregistration to Sentinel-2 or NAIP typically ranges from 3 to 10 m, which is reduced to 2–4 m after applying our MLP-based refinement, outperforming conventional polynomial and Rational Polynomial Coefficient (RPC) models. Further empirical evaluation demonstrates that consistent registration precision can be achieved across various terrain datasets; specifically, 30 m global digital elevation models (DEMs) yield comparable performance to that of high resolution (1–10 m) national or regional DEMs. The developed MLP approach provides a precise and reliable framework that transcends Landsat-specific tasks. It is readily adaptable to other registration applications requiring high-fidelity geometric alignment and consistent integration of heterogeneous image sources.</div></div>","PeriodicalId":100730,"journal":{"name":"ISPRS Open Journal of Photogrammetry and Remote Sensing","volume":"21 ","pages":"Article 100136"},"PeriodicalIF":7.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148657606","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"SMOL-MapSeg: Segmenting historical maps by legend-oriented prompting","authors":"Yunshuang Yuan,&nbsp;Frank Thiemann,&nbsp;Thorsten Dahms,&nbsp;Monika Sester","doi":"10.1016/j.ophoto.2026.100138","DOIUrl":"10.1016/j.ophoto.2026.100138","url":null,"abstract":"<div><div>Historical maps offer valuable insights into Earth’s surface changes but pose challenges for modern segmentation models due to inconsistent visual styles and symbols. While deep learning models like UNet and pre-trained foundation models perform well in domains such as autonomous driving and medical imaging, they struggle with the variability of historical maps, where similar concepts appear in diverse forms. To overcome this, we propose legend-oriented prompting, which provides an explicit image–label pair to guide models in identifying the target visual patterns shown in the prompt. This allows users to define and segment target patterns on demand, enabling flexible legend-based segmentation that mimics human map reading. Our approach replaces the prompt encoder of the Segment Anything Model (SAM) with the newly proposed prompting mechanism and fine-tunes it on historical maps, creating the SMOL-MapSeg (Show Me One Label<span><span>¹</span></span>) model. Unlike existing SAM-based fine-tuning methods that are class-agnostic or limited to fixed classes, SMOL-MapSeg is able to segment any target class across arbitrary datasets according to the class patterns defined by the prompt. Experiments show that a single SMOL-MapSeg model can accurately segment different classes, significantly outperforming baseline models. Moreover, it demonstrates strong generalization, even with limited training data, highlighting its potential for scalable and adaptable historical map analysis.</div></div>","PeriodicalId":100730,"journal":{"name":"ISPRS Open Journal of Photogrammetry and Remote Sensing","volume":"21 ","pages":"Article 100138"},"PeriodicalIF":7.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148951749","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"LB-DL method: A hybrid framework of Laplacian-Beltrami with deep learning models for spectral–spatial enhancement of wildfire monitoring remote sensing imagery","authors":"Mohammad Esmaeili,&nbsp;Dariush Abbasi-Moghadam,&nbsp;Alireza Sharifi","doi":"10.1016/j.ophoto.2026.100148","DOIUrl":"10.1016/j.ophoto.2026.100148","url":null,"abstract":"<div><div>Satellite imagery and remote sensing are suitable tools for monitoring and surveillance of the Earth. The use of this data to address various environmental challenges, from surveillance to risk management, is of great interest. In one of the applications of interest, satellite imagery is used for monitoring and assessing wildfires. However, its effectiveness is often limited by the spatial-spectral quality of the images, which also affects the performance of deep learning. This study proposes the LB-DL method, a hybrid framework that applies the Laplace-Beltrami (LB) operator to this application and integrates it with deep learning models to evaluate the improvement of image quality using LB. The LB operator exploits the inherent geometric structure of image manifolds for this improvement by suppressing noise while simultaneously enhancing critical spatial-spectral features, leading to more accurate mapping of burned areas. The method was evaluated using Sentinel-2 multispectral images of a wildfire in Uzbekistan and compared with conventional and state-of-the-art deep learning models. To test the robustness, realistic noise scenarios, additive, multiplicative, and dead pixels were simulated. Performance was measured using peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and classification accuracy. The results showed that LB-DL outperformed the baseline methods under the evaluated conditions, achieving PSNR improvements of up to 4.4 dB and SSIM increases of 0.13. Further validation on the hyperspectral datasets of Indian Pines and Salinas suggested promising generalizability, though additional validation on more diverse datasets and real-world conditions is needed. These findings highlight the potential of integrating geometry-based denoising with deep learning to enhance the reliability of remote sensing applications such as wildfire monitoring.</div></div>","PeriodicalId":100730,"journal":{"name":"ISPRS Open Journal of Photogrammetry and Remote Sensing","volume":"21 ","pages":"Article 100148"},"PeriodicalIF":7.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148951750","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}